SLAM for Indoor Mapping of Wide Area Construction Environments (2404.17215v1)

Abstract: Simultaneous localization and mapping (SLAM), i.e., the reconstruction of the environment represented by a (3D) map and the concurrent pose estimation, has made astonishing progress. Meanwhile, large scale applications aiming at the data collection in complex environments like factory halls or construction sites are becoming feasible. However, in contrast to small scale scenarios with building interiors separated to single rooms, shop floors or construction areas require measures at larger distances in potentially texture less areas under difficult illumination. Pose estimation is further aggravated since no GNSS measures are available as it is usual for such indoor applications. In our work, we realize data collection in a large factory hall by a robot system equipped with four stereo cameras as well as a 3D laser scanner. We apply our state-of-the-art LiDAR and visual SLAM approaches and discuss the respective pros and cons of the different sensor types for trajectory estimation and dense map generation in such an environment. Additionally, dense and accurate depth maps are generated by 3D Gaussian splatting, which we plan to use in the context of our project aiming on the automatic construction and site monitoring.

• The paper demonstrates a hybrid SLAM approach leveraging LiDAR and stereo cameras to achieve accurate 3D mapping in complex indoor construction sites.
• It compares LiDAR SLAM with visual SLAM, highlighting LiDAR's superior trajectory estimation and visual SLAM's ability to generate dense, detailed maps.
• The research underlines potential for future sensor fusion and semantic mapping to enhance automated monitoring and management of large indoor facilities.

SLAM for Indoor Mapping in Wide Area Construction Environments

Overview of Research

This paper investigates the application of the SLAM (Simultaneous Localization and Mapping) technique for data collection and mapping in large-scale indoor settings like factory halls and construction sites traditionally challenging due to their complex structural environments and the absence of Global Navigation Satellite System (GNSS) measures. The research involves the deployment of a mobile robotic system equipped with multiple stereo cameras and a LiDAR sensor aimed at capturing detailed 3D maps which are critical for efficient monitoring and management of such extensive facilities.

Methodology

Data Collection Instruments

The utilized system comprises an autonomous robot fitted with four stereo cameras and a 3D LiDAR scanner. These cameras provide a comprehensive 360-degree view, crucial for creating dense reconstructions of the environment. The 3D LiDAR sensor, notwithstanding its limited textural data capture capability compared to cameras, offers critical geometric and volumetric data, beneficial for creating accurate spatial mappings.

SLAM Approaches

The paper contrasts the performances of LiDAR-based and visual-based SLAM approaches:

• LiDAR SLAM: Showcased high accuracy in trajectory reconstructions thanks to its proficiency in handling larger measurement ranges. Its integration with IMU data aids in refining pose estimation, especially in dynamically changing or large-scale environments.
• Visual SLAM: Utilized dense mapping techniques facilitated by the stereo cameras. Although it generally produces denser maps favorable for detail-oriented tasks such as semantic segmentation, it faces challenges like scale drift common in stereo vision systems, particularly in extensive indoor arenas.

System Integration and Processing

Integration of data from both sensors using advanced algorithms such as 3D Gaussian splatting enhances depth perception and spatial resolution of the generated maps. Furthermore, these maps undergo subsequent optimization, providing detailed insights into environmental geometry and aiding in precise navigation and localization within complex structures.

Research Findings

Accuracy and Efficiency

The LiDAR SLAM provided sparse but highly accurate point clouds, essential for reliable mapping in large-scale environments. Visual SLAM generated denser maps, beneficial for detailed visual analysis but were affected by noise especially with distant objects affected by lighting variances or motion.

Enhanced Data Rendering

Applications of 3D Gaussian Splats in visual SLAM significantly improved the visual quality and completeness of depth maps allowing for the creation of realistic environmental visualizations and facilitating augmented reality applications more efficiently.

Implications and Future Directions

Practical Implications

The ability to generate accurate and detailed 3D maps of large indoor environments can significantly streamline operations in construction and manufacturing sectors. It reduces the time and labor costs associated with traditional manual monitoring and enables automated, precision-driven processes.

Theoretical Contributions and Speculations

This research contributes to the ongoing development and refinement of SLAM technologies, particularly in challenging environments lacking GNSS data. Future theoretical advancements may focus on hybrid models that integrate multiple sensor types to enhance both the accuracy and density of environmental mappings.

Scope for Future Research

Further research will explore the integration of LiDAR and visual data to harness the robust trajectory estimation of LiDAR with the high-density environmental details provided by visual SLAM. There is also potential for incorporating semantic information to update digital construction models automatically, aiding in more dynamic and context-aware building management systems.

Overall, this research underscores the substantial possibilities that SLAM technology presents for the future of automated indoor mapping and monitoring, suggesting avenues for both practical applications and fundamental SLAM research.